Molding machine

Abstract

This invention provides a molding machine capable of efficiently generating appropriate training data for use in machine learning of an AI that generates molding conditions. [Solution] The molding machine is a machine that performs a molding process in which a molding material is injected into a mold to form a molded product, and comprises a control device having a memory that stores condition data which is the execution condition of the molding process, and a state detection sensor that detects the state of the molding machine. Each time the molding process is performed, the control device stores the state data indicating the state of the molding machine detected by the state detection sensor in the memory in association with the condition data, and when the condition data stored in the memory is changed, it stores in the memory information indicating the change in the condition data and information indicating the change in the state data before and after the change in the condition data as training data for the AI ​​to learn.

Description

Technical Field

[0001] The present invention relates to a molding machine capable of generating learning data for AI.

Background Art

[0002] Conventionally, a molding machine that molds a molding material in a cavity of a mold to form a molded product is known. In such a molding machine, in order to form a high-quality molded product, it is necessary to appropriately set molding conditions. However, setting appropriate molding conditions requires a lot of knowledge and experience, and there is a problem that there are few operators who can perform the setting work.

[0003] Therefore, Patent Document 1 discloses a technique for supplementing the knowledge and experience of an operator by causing AI (Artificial Intelligence) that has been machine-learned in advance to generate molding conditions.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] To generate appropriate molding conditions for AI, a huge amount of learning data is required. However, until now, data extracted from a molding machine has been manually edited to generate learning data, and it cannot be said that an efficient generation method for appropriate generated data has been established.

[0006] The present invention has been made in view of the above circumstances, and an object thereof is to provide a molding machine capable of efficiently generating appropriate learning data used for machine learning of AI for generating molding conditions.

Means for Solving the Problems

[0007] To solve the above problems, the present invention provides a molding machine that performs a molding process in which a molding material is injected into a mold to form a molded product, comprising: a control device having a memory for storing condition data which is the execution condition of the molding process; and a state detection sensor for detecting the state of the molding machine, wherein each time the molding process is performed, the control device stores state data indicating the state of the molding machine detected by the state detection sensor in the memory in association with the condition data, and when the condition data stored in the memory is changed, it stores in the memory information indicating the change in the condition data and information indicating the change in the state data before and after the change in the condition data as training data for the AI ​​to learn. [Effects of the Invention]

[0008] According to the present invention, it is possible to efficiently generate appropriate training data for use in machine learning of an AI that generates molding conditions. [Brief explanation of the drawing]

[0009] [Figure 1] This is a side view of the injection molding machine according to this embodiment. [Figure 2] This is a hardware configuration diagram of an injection molding machine. [Figure 3] This diagram shows the neural network of a pre-trained model. [Figure 4] This is a flowchart of the molding process. [Figure 5] This is a flowchart of the training data generation process. [Figure 6] This is an example of data stored in memory. [Figure 7] This is an example of training data. [Modes for carrying out the invention]

[0010] The injection molding machine 10 according to the present invention will be described below with reference to the drawings. The injection molding machine 10 is a molding machine that injects molten resin (molding material) measured into a mold to form a molded product. However, the specific example of a molding machine is not limited to the injection molding machine 10, and may also be a die-casting machine that injects molten metal (molding material) into a mold to form a molded product.

[0011] [Configuration of injection molding machine 10] Figure 1 is a side view of the injection molding machine 10 according to this embodiment. Figure 2 is a hardware configuration diagram of the injection molding machine 10. As shown in Figures 1 and 2, the injection molding machine 10 mainly comprises a mold clamping device 20, an injection device 30, and a control device 60.

[0012] The mold clamping device 20 opens and closes the mold 21 and clamps it. Specifically, the mold clamping device 20 mainly comprises a fixed die plate 23 that supports the fixed side mold 22 and a movable die plate 25 that supports the movable side mold 24. The fixed side mold 22 and the movable side mold 24 are supported so as to face each other in the left-right direction (horizontal direction) of the injection molding machine 10.

[0013] The movable die plate 25 moves left and right along the tie bar 27 as the driving force of the mold opening / closing motor 28 is transmitted through the toggle link mechanism 26. When the movable die plate 25 moves to the left, the fixed mold 22 and the movable mold 24 separate. On the other hand, when the movable die plate 25 moves to the right, the fixed mold 22 and the movable mold 24 come into contact, forming a cavity (internal space) inside the mold 21. When further pressure is applied in the direction that moves the movable die plate 25 to the right, the fixed mold 22 and the movable mold 24 are clamped together.

[0014] The injection device 30 plasticizes, measures, and injects the molding material. In this embodiment, the injection device 30 is positioned opposite the clamping device 20 in the horizontal direction (to the right of the clamping device 20). The injection device 30 mainly comprises a heating cylinder 31, a screw 32, a hopper 33, and a hopper block 34.

[0015] The heating cylinder 31 is a cylindrical member extending in the left-right direction of the injection molding machine 10. The heating cylinder 31 mainly includes a resin passage 35 and a nozzle 36. Further, a band heater 39 for heating the heating cylinder 31 is attached to the outer peripheral surface of the heating cylinder 31. The band heater 39 is a so-called "thermocouple" that generates heat by receiving power supply from the control device 60, for example.

[0016] The resin passage 35 is a columnar space extending in the axial direction (longitudinal direction) inside the heating cylinder 31. The resin passage 35 communicates with the outside of the heating cylinder 31 (the cavity of the mold 21) through a nozzle 36 provided at the tip (front end) of the heating cylinder 31. In other words, the resin passage 35 is a space extending along the axial direction from the nozzle 36.

[0017] The screw 32 is a cylindrical member. On the outer peripheral surface of the screw 32, a groove extending spirally along the longitudinal direction of the screw 32 (hereinafter referred to as "spiral groove") is formed. The screw 32 is accommodated in the internal space of the heating cylinder 31 in a state where it can move in the left-right direction (hereinafter referred to as "forward and backward") and rotate in the injection molding machine 10. Further, the screw 32 in the heating cylinder 31 is configured to be replaceable. In other words, screws 32 with different specifications (for example, material, shape of the spiral groove, volume of the spiral groove) can be inserted into the heating cylinder 31.

[0018] The driving force of the injection motor 37 is transmitted to the screw 32 to cause it to move forward and backward, and the driving force of the metering motor 38 is transmitted to cause it to rotate. More specifically, when the injection motor 37 is rotated forward, the screw 32 moves (advances) toward the tip (i.e., the nozzle 36) of the heating cylinder 31. On the other hand, when the injection motor 37 is rotated backward, the screw 32 moves (retreats) toward the base end (i.e., the side opposite to the nozzle 36) of the heating cylinder 31.

[0019] Hereinafter, among the range within which the tip position of the screw 32 can reach inside the heating cylinder 31, the position closest to the nozzle 36 is denoted as the "forward limit", and the position farthest from the nozzle 36 is denoted as the "retreat limit". Also, the terms "clockwise rotation" and "counterclockwise rotation" of the injection motor 37 do not specify the absolute rotation direction, but only specify the relative relationship (that is, clockwise rotation and counterclockwise rotation are rotations in opposite directions).

[0020] The hopper 33 is a funnel-shaped member that stores granular resin as a raw material. The hopper block 34 is a member that supports the heating cylinder 31 and the hopper 33. The hopper 33 communicates with the resin passage 35 on the base end side from the tip of the heating cylinder 31 through the hopper block 34. The granular resin stored in the hopper 33 is supplied to the resin passage 35 of the heating cylinder 31 through the opening provided at the lower end. The granular resin used in this injection molding machine 10 is, for example, a so-called "pellet" formed in a cylindrical shape.

[0021] The injection device 30 rotates the injection motor 37 counterclockwise and rotates the metering motor 38, so that the screw 32 retreats while rotating. As a result, the pellets supplied through the hopper 33 are plasticized and filled (metered) into the resin passage 35 in front of the screw 32. Also, the injection device 30 rotates the injection motor 37 clockwise, so that the screw 32 advances. As a result, the molten resin filled in the resin passage 35 in front of the screw 32 is injected into the cavity of the mold 21 through the nozzle 36.

[0022] The hopper 33 is supplied with different types of resin (e.g., different degrees of plasticity) depending on the molded product. The particle size of the pellets supplied to the hopper 33 varies depending on the raw material supply device (not shown) that supplies the raw materials to the hopper 33. In addition to pellets, recycled resin may also be supplied to the hopper 33. Recycled resin refers to unwanted parts (runners) separated from the molded product, resin discharged (purged) from the heating cylinder 31, etc. The ratio of pellets and recycled resin supplied to the hopper 33 gradually changes during the molding process, which will be described later with reference to Figure 3.

[0023] [Configuration of the control device 60] As shown in Figure 2, the control device 60 comprises a CPU (Central Processing Unit) 61 and memory 62. The memory 62 is composed of, for example, ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disk Drive), or a combination thereof. The control device 60 performs the processing described later by having the CPU 61 read and execute program code stored in the ROM or HDD. RAM is used as a work area when the CPU 61 executes the program.

[0024] However, the specific configuration of the control device 60 is not limited to this and may be implemented using hardware such as an ASIC (Application Specific Integrated Circuit) or FPGA (Field-Programmable Gate Array).

[0025] The control device 60 controls the operation of the entire injection molding machine 10. More specifically, the control device 60 controls the mold opening / closing motor 28, injection motor 37, metering motor 38, band heater 39, and communication interface (IF) 68 based on various signals output from the rotary encoder 64, load cell 65 (pressure sensor), temperature sensor 66, and display input device 67.

[0026] The mold opening / closing motor 28, the injection motor 37, and the metering motor 38 are servo motors that generate driving force to open and close the mold 21, driving force to move the screw 32 forward and backward, and driving force to rotate the screw 32, respectively, according to the control of a servo amplifier (not shown).

[0027] The rotary encoder 64 is a sensor that detects the speed and tip position of the screw 32. More specifically, the rotary encoder 64 outputs pulse signals to the control device 60 corresponding to the rotation of the injection motor 37. The control device 60 then determines the speed of the screw 32 based on the number of pulse signals output per unit time. The control device 60 also determines the tip position of the screw 32 based on the cumulative value of the pulse signals.

[0028] The load cell 65 is a sensor that detects the pressure applied to the screw 32. More specifically, the load cell 65 outputs a pressure signal (voltage value) corresponding to the pressure applied to the screw 32 to the control device 60. The control device 60 then determines the pressure applied to the screw 32 based on the pressure signal output from the load cell 65.

[0029] The temperature sensor 66 detects the temperature of the heating cylinder 31 and outputs a temperature signal indicating the detected temperature to the control device 60. The control device 60 then determines the temperature of the heating cylinder 31 based on the temperature signal output from the temperature sensor 66.

[0030] The display input device 67 is a user interface that includes a display (display device) for displaying various information to be notified to the operator, and buttons, switches, dials, etc. (input devices) for receiving input operations from the operator. The display input device 67 may also include a touch panel superimposed on the display. The display input device 67 receives input operations from the operator and outputs an input signal corresponding to the received input operation to the control device 60.

[0031] The communication interface 68 is an interface for communicating with external devices (e.g., AI server 50, management terminal) via a communication network. The communication network consists of, for example, the internet, a public network, a wired LAN, a wireless LAN, or a combination thereof. The control device 60 transmits data to external devices and receives data from external devices via the communication interface 68. The functions of the AI ​​server 50 may also be implemented in the control device 60.

[0032] [Processing by AI Server 50] Figure 3 shows the neural network of the trained model 51. The AI ​​server 50 is implemented on a general-purpose computer such as a workstation or personal computer. The AI ​​server 50 implements AI (Artificial Intelligence) including the trained model 51. The AI ​​installed in the AI ​​server 50 processes input data and outputs output data. The AI ​​installed in the AI ​​server 50 also generates output data from input data using, for example, the neural network shown in Figure 3.

[0033] As shown in Figure 3, the neural network consists of an input layer L1 composed of multiple nodes I1, I2, and I3, a hidden layer L2 composed of multiple nodes H1, H2, H3, and H4, and an output layer L3 composed of multiple nodes O1, O2, and O3. In the example in Figure 3, the number of nodes in the input layer L1 and the output layer L3 are the same, but the number of nodes in the input layer L1 and the output layer L3 may be different. Furthermore, the neural network may have multiple hidden layers L2. In addition, Figure 3 shows a fully connected neural network in which the multiple nodes constituting each layer L1, L2, and L3 are connected to all nodes in adjacent layers, but the structure of the neural network is not limited to this.

[0034] The trained model 51 is generated by inputting multiple training data, including input data and ground truth data, into the pre-trained model (hereinafter referred to as the "pre-trained model"). Input data refers to the data that is input to the pre-trained model. Ground truth data refers to the data that should be output when the input data is input. By inputting multiple training data into the pre-trained model, the neural network is optimized to become the trained model 51. The training data is generated, for example, based on the results of experiments, simulations, or molding processes performed in the injection molding machine 10. This process is an example of a training process that adjusts the weight coefficients and biases of each node so that when input data is input to the input layer L1, ground truth data is output from the output layer L3.

[0035] Furthermore, the learning process may be performed not only on the pre-trained model but also on the trained model 51. In addition, the AI ​​may learn not only using the input data and ground truth data actually used in this invention, but also with general-purpose learning data. Moreover, the learning process performed by the AI ​​may be not only "supervised learning" in which input data and ground truth data are input, but also "unsupervised learning" in which ground truth data is not input, or it may be reinforcement learning or transfer learning, etc.

[0036] Furthermore, the AI ​​server 50 generates and outputs output data by inputting input data into the neural network of the trained model 51. This process is an example of a generation process in which the input data input to the input layer L1 is processed using weight coefficients and biases that have been adjusted in advance for each node, and output from the output layer L3.

[0037] [Explanation of conditional data] Memory 62 stores condition data. Condition data is the execution conditions for the molding process described later (hereinafter referred to as "molding conditions"). In the molding process shown in Figure 3, the control device 60 operates the injection molding machine 10 according to the condition data stored in the condition storage area of ​​memory 62. The condition data according to this embodiment includes multiple items (for example, injection speed, injection pressure, injection stroke, holding pressure time, holding pressure, metering speed, rotation speed, screw back pressure, heater temperature). However, the items included in the condition data are not limited to these.

[0038] The item "Injection Speed" is the forward speed (mm / s) of the screw 32 during the injection process. The item "Injection Pressure" indicates the maximum pressure (MPa) applied to the screw 32 during the injection process. The item "Injection Stroke" is the forward distance (mm) of the screw 32 during the injection process. In other words, during the injection process (S12), the control device 60 drives the injection motor 37 so that the forward speed of the screw 32 approaches the item "Injection Speed" and the pressure applied to the screw 32 does not exceed the item "Injection Pressure," and stops the screw 32 at a point where it has advanced by the item "Injection Stroke."

[0039] The item "Holding time" indicates the execution time of the holding process. The item "Holding pressure" indicates the maximum pressure (MPa) applied to the screw 32 during the holding process. In other words, during the holding process (S13), the control device 60 drives the injection motor 37 so that the pressure applied to the screw 32 does not exceed the "Holding pressure" until the "Holding time" has elapsed.

[0040] The item "metering speed" is the retraction speed (mm / s) of the screw 32 during the metering process. The item "rotational speed" is the rotational speed (rpm) of the screw 32 during the metering process. The item "screw back pressure" is the pressure (MPa) applied to the screw 32 during the metering process. In other words, the control device 60 drives the injection motor 37 and the metering motor 38 during the metering process (S14) so ​​that the retraction speed of the screw 32 approaches the item "metering speed", the rotational speed of the screw 32 approaches the item "rotational speed", and the back pressure applied to the screw 32 does not exceed the item "screw back pressure".

[0041] The item "heater temperature" is the temperature (°C) of the heating cylinder 31. In other words, the control device 60 controls the power supply to the band heater 39 so that the temperature of the heating cylinder 31 approaches the item "heater temperature" during the molding process. More specifically, when the temperature of the heating cylinder 31 is lower than the item "heater temperature", the control device 60 increases the power supply time to the band heater 39 as the temperature of the heating cylinder 31 decreases, and decreases the power supply time to the band heater 39 as the temperature of the heating cylinder 31 increases.

[0042] The control device 60 stores new molding condition data in the condition storage area according to instructions from the operator via the display input device 67, or according to instructions from an external device (e.g., AI server 50, management terminal not shown) via the communication IF 68. The control device 60 also modifies the molding condition data stored in the condition storage area (i.e., changes the values ​​of each item) according to instructions from the operator via the display input device 67, or according to instructions from an external device (e.g., AI server 50, management terminal not shown) via the communication IF 68. The display input device 67 and the communication IF 68 are interfaces for instructing changes to the molding condition data.

[0043] [Molding process] Figure 4 is a flowchart of the molding process. The molding process involves injecting molten resin, which has been filled into the heating cylinder 31, into the cavity of the clamped mold 21, in accordance with the condition data stored in the condition memory area, thereby forming a molded product. At the start of the molding process, the mold 21 is open, the molten resin to be injected next is metered into the space in front of the screw 32 of the heating cylinder 31, and the helical groove of the screw 32 on the tip side of the hopper 33 is filled with resin (pellets, resin in the process of plasticization, molten resin).

[0044] First, the control device 60 rotates the mold opening / closing motor 28 to close and clamp the mold 21 (S11). This creates a cavity inside the mold 21. The process in step S11 is an example of a mold clamping process.

[0045] Next, after the mold clamping process (S11) is completed, the control device 60 advances the screw 32 by rotating the injection motor 37 in the forward direction according to the molding conditions "injection speed", "injection pressure", and "injection stroke" (S12). As a result, the molten resin metered in the area in front of the screw 32 within the heating cylinder 31 is injected into the cavity of the mold 21. The process in step S12 is an example of the injection process.

[0046] Next, after the injection process (S12) is completed, the control device 60 applies pressure to the resin injected into the mold 21 by rotating the injection motor 37 according to the molding conditions "holding time" and "holding pressure" (S13). The process in step S13 is an example of the holding pressure process.

[0047] Next, after the holding pressure process (S13) is completed, the control device 60 rotates the injection motor 37 and the metering motor 38 according to the molding conditions "metering speed", "rotation speed", and "screw back pressure", thereby rotating and retracting the screw 32 (S14). As a result, the granular resin supplied to the heating cylinder 31 through the hopper 33 is plasticized and metered into the space in front of the screw 32 in the heating cylinder 31. The process in step S14 is an example of a metering process.

[0048] Next, after the weighing process (S14) is completed, the control device 60 rotates the mold opening / closing motor 28 to open the mold 21, and causes a robot arm (not shown) to remove the molded product from the opened mold 21 (S15). The process in step S15 is an example of the removal process.

[0049] Furthermore, the control device 60 continuously controls the power supply to the band heater 39 so that the temperature of the heating cylinder 31 approaches the item “heater temperature” during the execution of the molding process.

[0050] Furthermore, the control device 60 acquires status data (monitor data) during the execution of the molding process. Status data is data indicating the state of the injection molding machine 10 performing the molding process. The status data includes, for example, multiple items (e.g., screw forward speed, maximum pressure during injection, actual stroke, maximum pressure during holding pressure, screw retraction speed, energization rate). However, the items included in the condition data are not limited to these.

[0051] The item “Screw forward speed” is the forward speed of the screw 32 detected by the rotary encoder 64 during the injection process (S12) (for example, the forward speed at a specific position). The item “Maximum pressure during injection” is the maximum value of the pressure detected by the load cell 65 during the injection process (S12). The item “Actual stroke” is the forward distance of the screw 32 detected by the rotary encoder 64 during the injection process (S12) and the holding pressure process (S13).

[0052] The item “Maximum pressure during holding pressure” is the maximum pressure detected by the load cell 65 during the holding pressure process (S13). The item “Screw retraction speed” is the retraction speed of the screw 32 detected by the rotary encoder 64 during the metering process (S14) (for example, the retraction speed at a specific position). The item “Energization rate” is the ratio of the power supply time to the band heater 39 to the execution time of the molding process.

[0053] The rotary encoder 64, which detects the items “screw forward speed,” “actual stroke,” and “screw retraction speed,” the load cell 65, which detects the items “maximum pressure during injection” and “maximum pressure during holding pressure,” and the control device 60, which detects the item “energy rate,” are examples of state detection sensors for detecting the state of the injection molding machine 10. However, specific examples of state detection sensors are not limited to these.

[0054] [Training data generation process] Figure 5 is a flowchart of the training data generation process. Figure 6 is an example of data stored in memory 62. Figure 7 is an example of training data. The training data generation process is a process in which the injection molding machine 10 generates training data for machine learning on the AI ​​server 50 during the process of repeatedly executing the molding process. In Figures 6 and 7, the symbol "b (before)" is added to the data and items before the condition data is changed, and the symbol "a (after)" is added to the data and items after the condition data is changed.

[0055] First, the control device 60 executes the molding process according to the condition data b stored in the condition storage area (S21). Then, as shown in Figure 6(A), the control device 60 associates the state data b acquired in the molding process of step S21 with the condition data b stored in the condition storage area and stores it in the first state storage area of ​​memory 62 (S22). The control device 60 then repeatedly executes the processes of steps S21 to S22 until the condition data b in the condition storage area is changed (S23: No).

[0056] The control device 60 stores, for example, the state data b1, b2, and b3 acquired in the most recent L molding processes in the first state storage area, associating them with the condition data b. That is, each time a molding process is acquired, the control device 60 deletes the state data with the oldest acquisition date and time from the state data b1, b2, and b3 stored in the first state storage area and stores the newly acquired state data. L is an integer of 2 or more (3 in this embodiment). Also, before the condition data in the condition storage area is changed, the condition data b in the first state storage area is identical to the condition data in the condition storage area.

[0057] Next, if the condition data b stored in the condition memory area is changed (S23:Yes), the control device 60 executes the molding process according to the changed condition data a (S24). Next, if the molding process after the change in condition data (S24) is less than N times (S25:No), the control device 60 executes the molding process again (S24) without executing the processes from step S26 onwards. In other words, the control device 60 does not store the condition data acquired in the 1st to (N-1)th molding processes after the change in condition data in the second state memory area. N is an integer of 2 or more. N is set to the number of times until the molding process according to the changed condition data becomes stable (for example, 10).

[0058] Next, if the control device 60 executes the Nth or subsequent molding process after the change in the condition data (S25:Yes), as shown in Figure 6(B), it associates the state data a acquired in the molding process of step S24 with the condition data stored in the condition storage area (i.e., the changed condition data a) and stores it in the second state storage area of ​​the memory 62 (S26).

[0059] The control device 60 then repeatedly executes the processes in steps S24 to S26 until the number of state data stored in the second state storage area reaches M (S27: No). That is, the control device 60 stores the state data a1, a2, and a3 acquired in the Nth to (N+M)th molding processes after the change in the condition data in the second state storage area, associating them with the condition data a stored in the condition storage area. M is an integer of 2 or more (3 in this embodiment).

[0060] Next, when the control device 60 stores M state data a1, a2, and a3 in the second state memory area (S27: Yes), it generates training data using the condition data b and state data b1, b2, and b3 stored in the first state memory area, and the condition data a and state data a1, a2, and a3 stored in the second state memory area (S28). The specific method for generating the training data will be described later. Then, the control device 60 stores the generated training data in the training data storage area of ​​the memory 62.

[0061] Furthermore, when the control device 60 generates learning data (S28), it deletes the condition data b and state data b1, b2, and b3 stored in the first state memory area, stores the condition data a and state data a1, a2, and a3 stored in the second state memory area as condition data b and state data b1, b2, and b3 in the first state memory area, and deletes the condition data a and state data a1, a2, and a3 stored in the second state memory area. The control device 60 then accumulates learning data in the learning data memory area by repeatedly executing the learning data generation process.

[0062] The control device 60 includes in the training data information that indicates changes in condition data and information that indicates changes in state data before and after the change in condition data. Information that indicates changes in condition data is, for example, the absolute value of each item of condition data a and b, the difference (ba) between each item of condition data a and b, or the ratio (a / b) of each item of condition data. Similarly, information that indicates changes in state data is, for example, the absolute value of state data a and b, the difference (ba) between state data a and b, or the ratio (a / b) of state data a and b.

[0063] The following explains the process of generating training data using the case where the "holding time" item in the condition data is changed as an example. Furthermore, an example of "difference" is explained as information indicating changes in the condition data and state data. Additionally, Figure 7 illustrates only a portion of the state data items.

[0064] First, the differences between each item of condition data b stored in the first state memory area and each item of condition data a stored in the second state memory area—namely, "injection speed (ba)", "injection pressure (ba)", "injection stroke (ba)", "holding time (ba)", "holding pressure (ba)", "metering speed (ba)", "rotation speed (ba)", "screw back pressure (ba)", and "heater temperature (ba)"—are included in the training data. In Figure 7, these are collectively referred to as condition data (ba).

[0065] Furthermore, the control device 60 includes the item-by-item difference between one of the multiple state data b1, b2, and b3 in the first state memory area (i.e., before the change of the condition data) (typically, state data b3 immediately before the change of the condition data) and one of the multiple state data a1, a2, and a3 in the second state memory area (i.e., after the change of the condition data) (typically, state data a1 immediately after the change of the condition data) in the training data.

[0066] As another example, the control device 60 may calculate the average value Ave(b) for each item of multiple state data b1, b2, and b3 in the first state memory area (i.e., before the condition data is changed). Similarly, the control device 60 may calculate the average value Ave(a) for each item of multiple state data a1, a2, and a3 in the second state memory area (i.e., after the condition data is changed). The control device 60 may then include the difference (Ave(b)-Ave(a)) between the average value before the condition data is changed and the average value after the condition data is changed for each item of the state data in the training data.

[0067] Furthermore, the control device 60 may, for example, determine whether or not to include the detected value of the state detection sensor in the learning data from among the multiple items included in the state data, based on the correspondence shown in Figure 6(C). Figure 6(C) is a table that stores the items detected by the state detection sensor (middle column) and the items not detected by the state detection sensor (right column) from among the multiple items of the state data, corresponding to each of the multiple items of the condition data (left column). The table in Figure 6(C) is pre-stored in the memory 62. Also, it is sufficient if either the items detected by the state detection sensor or the items not detected by the state detection sensor are stored. Furthermore, only representative items are shown in the table in Figure 6(C).

[0068] As an example, the control device 60 calculates the difference "S(ba)" and "HP(ba)" between the values ​​detected by the state detection sensor before and after the change in the condition data for the items "Actual Stroke S" and "Maximum Pressure IP during Holding Pressure" stored in the middle column of the table in Figure 6(C) in association with the condition data item "Holding Pressure Time," and includes these in the learning data. On the other hand, the control device 60 sets the items "Maximum Pressure IP during Injection" and "Energy Rate ER," which are not stored in the middle column of the table in Figure 6(C) in association with the condition data item "Holding Pressure Time," to 0 and includes them in the learning data.

[0069] As another example, the control device 60 calculates the difference "S(ba)" and "HP(ba)" between the values ​​detected by the state detection sensor before and after the change in the condition data for the items "actual stroke S" and "maximum pressure HP during holding pressure," which are not the items "maximum pressure IP during injection" and "energy rate ER" stored in the right column of the table in Figure 6(C) in association with the condition data item "holding pressure time," and includes them in the learning data. On the other hand, the control device 60 sets the items "maximum pressure IP during injection" and "energy rate ER," which are stored in the right column of the table in Figure 6(C) in association with the condition data item "holding pressure time," to 0 and includes them in the learning data.

[0070] Furthermore, the control device 60 may store the learning data generated in step S29 in a learning data storage area, associating it with other information of the injection molding machine 10. For example, the control device 60 may store the learning data associating it with the identifier of the mold 21 (Figure 7(A)). As another example, the control device 60 may store the learning data associating it with the identifier of the molding material (pellet) (Figure 7(B)). As yet another example, the control device 60 may store the learning data as association with the identifier of the operator (not shown) who changed the condition data in step S23.

[0071] An identifier is information that uniquely identifies the mold 21, the molding material, and the operator. That is, the control device 60 may store learning data in the learning data storage area for each type of mold 21 used in the molding process in steps S21 and S24, each type of molding material (pellets) used in the molding process in steps S21 and S24, or for each operator whose condition data was changed in step S23. In addition, it is assumed that the same mold 21 and the same molding material are used in the molding process (S21, S24) which is repeatedly executed in a single learning process.

[0072] Furthermore, the training data stored in the training data storage area is used for training the AI ​​server 50. The training data may be input to the AI ​​server 50 via a portable storage medium (e.g., a USB memory stick), or the control device 60 may transmit it to the AI ​​server 50 via the communication IF 68. The training data input to the AI ​​server 50 may also include identifiers for the corresponding mold 21, molding material identifiers, or operator identifiers. In addition, the training data stored in the training data storage area may be shuffled before being input to the AI ​​server 50.

[0073] Note that the information showing changes in state data is an example of input data, and the information showing changes in condition data is an example of correct data. In other words, the learning data according to this embodiment shows how to change the condition data (for example, increase the holding pressure time by 5 seconds) when you want to change the state data (for example, increase the actual stroke by 0.5 mm). In addition, the identifier of the mold 21, the identifier of the molding material, or the identifier of the operator may also be included in the input data.

[0074] Furthermore, the AI ​​server 50 (trained model 51), which has been trained using the training data generated by the injection molding machine 10, may, for example, suggest changes to the condition data in order to bring the state data closer to the ideal state. The control device 60 may, for example, input information indicating the deviation (change) between the actual value and the ideal value of the state data to the AI ​​server 50 as input data. The control device 60 may then obtain from the AI ​​server 50 as output data how to change the condition data (for example, reduce the holding pressure time by 3 seconds) in order to bring the actual value of the input data closer to the ideal value (for example, reduce the actual stroke by 0.5 mm).

[0075] [Effects of the Embodiment] According to the above embodiment, learning data can be automatically accumulated during the actual operation of the injection molding machine 10 (for example, during the process of adjusting molding conditions to bring the molded product closer to the desired quality). This makes it possible to efficiently generate appropriate learning data to be used for machine learning of the AI ​​server 50 that generates the molding conditions. Furthermore, by using changes in condition data as a trigger to store condition data and state data as a set, the data capacity of the learning data can be reduced compared to storing all the state data acquired each time injection molding is performed.

[0076] Furthermore, according to the above embodiment, state data acquired in (N-1) shaping processes immediately after the condition data is changed is not included in the training data. This prevents the AI ​​server 50 from learning with unstable state data immediately after the condition data is changed. However, if N=1, the state data immediately after the condition data is changed may be included in the training data.

[0077] Furthermore, according to the above embodiment, training data is generated using the average value of state data obtained under L molding conditions before the change in condition data and the average value of state data obtained under M molding conditions after the change in condition data. This makes it possible to equalize the detection variability by the state detection sensor. However, the specific example of the representative value is not limited to the mean (e.g., arithmetic mean, geometric mean, harmonic mean), but may also be the median, mode, etc. Also, L and M may be the same value or different values. Also, L=M=1.

[0078] However, the state data used to generate training data is not limited to the state data b1, b2, and b3 immediately before the condition data is changed; any condition data from before the condition data was changed is acceptable. Similarly, the state data used to generate training data is not limited to the state data a1, a2, and a3 immediately after the condition data is changed; any condition data from after the condition data has been changed (or from the Nth time onward after the condition data has been changed) is acceptable.

[0079] Furthermore, according to the above embodiment, depending on the item in the condition data that has been changed, the value of the state data to be included in the training data is switched between being the detected value of the state detection sensor or being forcibly set to 0. This prevents items among the multiple items included in the state data that are only loosely related to the item in the changed condition data (e.g., injection velocity) (e.g., current rate) from being affected by the detection variability of the state detection sensor. However, all items of the state data may also be included in the training data as the detected value of the state detection sensor.

[0080] Furthermore, according to the above embodiment, learning data is accumulated for each piece of information about the injection molding machine 10 (for example, the type of mold 21, the type of molding material, and the operator). This allows the AI ​​server 50 to perform machine learning while taking into account the state of the injection molding machine 10. In particular, by having the AI ​​server 50 perform machine learning with learning data for each operator, it is possible to build an expert system that takes into account the habits of operators who change the molding conditions.

[0081] Furthermore, according to the above embodiment, by including the difference (or ratio) of each data before and after the change in the condition data in the training data, the data capacity of the training data can be reduced compared to the case where the absolute values ​​of each data before and after the change are included. This makes it possible to generate training data even in an injection molding machine 10 with a small memory capacity 62. However, the training data may also include the absolute values ​​of each data before and after the change in the condition data.

[0082] The embodiments described above are illustrative for explaining the present invention and are not intended to limit the scope of the invention to those embodiments only. Those skilled in the art can implement the present invention in various other forms without departing from the spirit of the invention. [Explanation of symbols]

[0083] 10…Injection molding machine, 20…Clamping device, 21…Mold, 22…Fixed mold, 23…Fixed die plate, 24…Movable mold, 25…Movable die plate, 26…Toggle link mechanism, 27…Tie bar, 28…Mold opening / closing motor, 30…Injection device, 31…Heating cylinder, 32…Screw, 33…Hopper, 34…Hopper block, 35…Resin passage, 36…Nozzle, 37…Injection motor, 38…Measuring motor, 39…Band heater, 50…AI server, 51…Trained model, 60…Control device, 61…CPU, 62…Memory, 64…Rotary encoder, 65…Load cell, 66…Temperature sensor, 67…Display input device, 68…Communication interface

Claims

1. In a molding machine that performs a molding process in which molding material is injected into a mold to form a molded product, A control device having a memory that stores condition data which is the execution condition of the molding process, The system includes a state detection sensor for detecting the state of the molding machine, The control device is Each time the molding process is performed, the state data indicating the state of the molding machine detected by the state detection sensor is stored in the memory in association with the condition data. A molding machine characterized in that, when the condition data stored in the memory is changed, information indicating the change in the condition data and information indicating the change in the state data before and after the change in the condition data are stored in the memory as training data for the AI ​​to learn.

2. In the molding machine according to claim 1, The molding machine is characterized in that the control device includes in the learning data information indicating the change between the state data acquired in the molding process before the change in the condition data and the state data acquired in the N (where N is an integer of 2 or more)th molding process after the change in the condition data.

3. In the molding machine according to claim 1, The molding machine is characterized in that the control device includes in the learning data information showing the change between the representative value of the state data obtained in L (where L is an integer of 2 or more) molding processes before the change of the condition data and the representative value of the state data obtained in M ​​(where M is an integer of 2 or more) molding processes after the change of the condition data.

4. In the molding machine according to claim 1, The aforementioned condition data and the aforementioned state data each include multiple items, The memory pre-stores, in association with each of the multiple items of the condition data, the items from the multiple items of the state data that are detected by the state detection sensor. The molding machine is characterized in that the control device detects, using the state detection sensor, an item stored in association with the changed item of the condition data from among a plurality of items of the state data and includes it in the learning data.

5. In the molding machine according to claim 1, The aforementioned condition data and the aforementioned state data each include multiple items, The memory pre-stores, in association with each of the multiple items of the condition data, the items from the multiple items of the state data that are not detected by the state detection sensor. The molding machine is characterized in that the control device detects items from among the multiple items of the state data that are not stored in association with the changed item of the condition data using the state detection sensor and includes them in the learning data.

6. In the molding machine according to claim 1, The molding machine is characterized in that the control device stores the learning data in the memory for each operator who changes the type of mold, the type of molding material, or the condition data.

7. In the molding machine according to claim 1, The information indicating the change in the condition data is the absolute value of the condition data before and after the change, the difference between the condition data before and after the change, or the ratio of the condition data before and after the change. The molding machine is characterized in that the information indicating the change in the state data is the absolute value of the state data before and after the change in the condition data, the difference between the state data before and after the change in the condition data, or the ratio of the state data before and after the change in the condition data.

Images